The Listening Cycle · episode 6
The group that vanished
Composite scenario. Unofficial analysis for illustration. Not a named firm.
The organisation
A UK regional grocer — 2,200 colleagues, 16 stores and a distribution centre. They field the annual census in Culture Amp. We do not have a Culture Amp connector; the cycle is their Excel into analysis. Last May the board pack showed Store 14 night-fill as 41% (n=9). A store manager could name the nine. Trust in the listening died on the shop floor before the actions were read.
The constraint
A score is blank because the cell is too small, and because leaving it as the only blank cell lets the room reconstruct it from the total. Complementary suppression is a professional obligation, not a unique claim. Do not market k-anonymity — the privacy page refuses that guarantee. n-gate 30 on reported cuts. Distinctive comments do not ship.
Three decisions
1. n=9 is not a store score. It is nine people
Store 14 night-fill, n=9, 41%. Anyone who works the shift can do the arithmetic. The cell is blank. The pack says so, in a footnote the board can read: ‘blank where n < 30’.
2. One blank cell is a subtraction problem
If Store 14 is the only store missing, region minus the other fifteen is Store 14. So Store 7 (n=34 — it would otherwise show) goes blank too. That is complementary suppression. It is why two stores vanish, not one. It is not a k-anonymity guarantee.
3. Four comments do not ship
27 items, 11 dimensions, n=211 at the DC, four comments suppressed because a turn of phrase names a manager. Verbatim or it does not ship; distinctive or it does not ship. The suppression log is for the People director, not a slide of ‘colourful quotes’.
The cycle
| Channel | When | Who fields |
|---|---|---|
| Culture Amp census (theirs) Head of People | March, all stores + DC | Culture Amp — keep it. Excel export into analysis; no connector |
| Board pack that blanks two stores People director + retail ops | May, company master | Analysis here. Complementary suppression on. Export blocked if a sub-30 cell is visible |
| Store-manager huddles Regional managers | June, only stores that cleared n-gate | Offline — their own cut, not a dump of the nine at Store 14 |
- Sep — Sign the cycle. Write the rule on one slide: n-gate 30; complementary suppression; no k-anonymity claim.
- Oct–Feb — Peak trading. No pulses. Exit conversations continue.
- Mar — They field in Culture Amp. We take the Excel.
- Apr — Apply n-gate 30. Store 14 night-fill blanks. Store 7 blanks with it so the board cannot subtract. Four comments out.
- May — Board pack in their master. The blank cells have a sentence, not a grey box with no legend. ‘Why is Store 14 blank?’ is the meeting, and it has an answer.
- Jun — Huddles at stores that cleared the gate. Store 14 gets a conversation, not a dashboard.
- Jul — Four company-level actions as WOOP. Do not set a Store 14 target on nine people.
- Aug — Review the cycle. The suppression log is an artefact, not a marketing claim.
What we will not ask
- A k-anonymity guarantee the privacy page refuses.
- A store score on n=9.
- One blank cell the board can subtract.
- A distinctive comment that names a manager.
- A Culture Amp connector we do not have.
90-minute workshop
20 min · Name the nine
Store 14, 41%, n=9. The room has to admit they could name them. Then we blank the cell. That is the point.
25 min · Subtract the region
Blank only Store 14, then watch someone reconstruct it from the total. Blank Store 7 with it. Complementary suppression is why two stores vanish.
25 min · Read the sentence on the slide
‘Blank where n < 30; a second cell suppressed so the first cannot be subtracted. This is not a k-anonymity guarantee.’ The board sees that sentence.
20 min · What Store 14 gets instead
A conversation, not a dashboard. Four company actions that do not require the nine to be a KPI. Sign the cycle or stop.